Search

In this paper, we propose a novel paper currency defect detection algorithm using quaternion uniform strength. We first build paper currency image preprocessing integration framework which includes intensity balancing, paper currency location, and geometric correction. We then propose a global–local paper currency image registration algorithm by moving key areas within certain range which can eliminate the false difference effectively. Finally, the quaternion uniform strength is calculated by using quaternion convolution edge detector. The defect degree of paper currency is determined by using the quaternion uniform color difference. The proposed algorithm is tested using different datasets from five countries: CNY, USD, EUR, VND, and RUB. The experimental results demonstrate that the proposed algorithm yields better results than the existing state-of-the-art paper currency defect detection techniques. The demo of the proposed paper currency defect detection algorithm will be publicly available. [ABSTRACT FROM AUTHOR]

This paper presents a convolutional neural network (CNN) to classify between the conventionally and organically cultivated Memecik varieties of green olives. The image forming method called the rising paper chromatography is utilized in preparing the images of Memecik varieties of green olives for CNN. In the rising chromatography method, 20, 30, and 40% sample concentrations were determined as the suitable concentrations for both organic and conventional olives. The concentrations of AgNO3 and FeSO4 were determined as 0.25, 0.5, 0.75 and 1% for both conventional and organic samples. The visual differences used for differentiation of different types of Memecik green olives are usually determined according to the regional color differences, the vivid color occurrence, the width and the frequency of bowl occurrence, the thin line, and the picks at drop zone by the expert assessors. The testing results in this study verified the effectiveness of the CNN methodology in differentiating between the organically and conventionally cultivated Memecik green olives. The newly designed neural network achieved 100% accuracy. Furthermore, this high accuracy achieved by CNN might suggest that it can be effectively used in place of the expert assessors. [ABSTRACT FROM AUTHOR]

Safety helmet wearing detection based on video surveillance is an important means of safety monitoring in many industrial scenes. The training of safety helmet wearing detection models requires large and well-labeled dataset. However, the incidence of security violations is relatively low, which results in insufficient samples for training deep detection models. Safety helmet wearing detection is a common requirement in many scenarios such as construction sites, substations, and factory workshops. Aggregating data from multiple companies for model training would improve the performance of the detection model. Traditional centralized training methods are not feasible because aggregating data in centralized locations (such as the cloud) can raise concerns about data privacy and the high cost of data communication and storage. This paper proposes FedSH, a novel cloud-edge-based federated learning framework, which learns a shared global safety helmet wearing detection model in the cloud from multiple companies at the network edges and achieves data privacy protection by keeping company data locally. In addition, this paper designs reweighting mechanisms and applies transfer learning to address class imbalance and non-IID problems in the training data, so as to obtain an accurate and personalized detection model. Extensive experiments have been conducted on real surveillance video datasets. The experimental results demonstrate that FedSH outperforms the existing widely used federated learning methods with an accuracy improvement of at least 3.4%; the reduction in accuracy is within the range of 5% compared with centralized learning methods. FedSH effectively achieves a good balance between model performance, privacy protection, and communication efficiency. [ABSTRACT FROM AUTHOR]

Natural disasters can be unpredictable and catastrophic. Even after the event, the repercussions are prolonged due to the incompetence of disaster management strategies. To mitigate the effects of a natural hazard, disaster management teams have to rapidly come forth with innovative plans of action. To contain the damage that a disaster causes, the response time, and preparedness of a disaster management team is crucial. Having a forewarning about the nature of the disaster can prove to be beneficial for the management team. Remote/unreachable areas such as deep parts of the forests, far-flung rural areas, oceans and other hard to reach locations are at a higher risk of receiving poorer aid and response due to the lack of communication and connectivity with the rest of the world. Our paper provides a solution for this particular problem by suggesting UAV/Drones for surveillance and monitoring in these inaccessible, disaster struck areas. The drones not only monitor the situation but constantly click pictures and send them to a base station, where the images are used to collect insights about the type of disaster that has to be dealt with. The paper proposes a deep learning model based on feature concatenation for classification of disasters which can be deployed at the base station, and can receive data in the form of images from a UAV/drone hovering over the affected place. The proposed model is efficient and able to achieve a higher accuracy as compared to the leading CNN models and closely related recent works as well. [ABSTRACT FROM AUTHOR]

In order to realize the intelligent recognition and statistics of hydrogenation catalyst image information, this paper presents a new method to judge the active phase by image recognition, which is different from traditional methods. Firstly, considering that hydrogenation catalyst image targets are small and easy to stack, the feature extraction network in the CenterNet model is optimized by adding the multi-feature fusion module to improve the accuracy of the network in edge positioning. Secondly, according to the linear shape of the hydrogenation catalyst, the mask branch is added to the CenterNet model to train the hydrogenation catalyst stripes with unclear target to reduce the leakage rate of the hydrogenation catalyst. The experimental results show that the detection accuracy of the improved CenterNet network is 91 % , 7 % higher than that of the original one, with a decline in detection rate by 12 % . The method proposed in this paper can accurately identify and segment the hydrogenation catalyst in the electron microscope image, which can provide technical support for the statistics and analysis of the hydrogenation catalyst image. [ABSTRACT FROM AUTHOR]

Medical image segmentation is crucial in using digital images for disease diagnosis, particularly in post-processing tasks such as analysis and disease identification. Segmentation of magnetic resonance imaging (MRI) and computed tomography images pose distinctive challenges attributed to factors such as inadequate illumination during the image acquisition process. Multilevel thresholding is a widely adopted method for image segmentation due to its effectiveness and ease of implementation. However, the primary challenge lies in selecting the optimal set of thresholds to achieve accurate segmentation. While Otsu's between-class variance and Kapur's entropy assist in identifying optimal thresholds, their application to cases requiring more than two thresholds can be computationally intensive. Meta-heuristic algorithms are commonly employed in literature to calculate the threshold values; however, they have limitations such as a lack of precise convergence and a tendency to become stuck in local optimum solutions. In this paper, we introduce an improved chameleon swarm algorithm (ICSA) to address these limitations. ICSA is designed for image segmentation and global optimization tasks, aiming to improve the precision and efficiency of threshold selection in medical image segmentation. ICSA introduces the concept of the "best random mutation strategy" to enhance the search capabilities of the standard chameleon swarm algorithm (CSA). This strategy leverages three distribution functions—Levy, Gaussian, and Cauchy—for mutating search individuals. These diverse distributions contribute to improved solution quality and help prevent premature convergence. We conduct comprehensive experiments using the IEEE CEC'20 complex optimization benchmark test suite to evaluate ICSA's performance. Additionally, we employ ICSA in image segmentation, utilizing Otsu's approach and Kapur's entropy as fitness functions to determine optimal threshold values for a set of MRI images. Comparative analysis reveals that ICSA outperforms well-known metaheuristic algorithms when applied to the CEC'20 test suite and significantly improves image segmentation performance, proving its ability to avoid local optima and overcome the original algorithm's drawbacks. Medical image segmentation is essential for employing digital images for disease diagnosis, particularly for post-processing activities such as analysis and disease identification. Due to poor illumination and other acquisition-related difficulties, radiologists are especially concerned about the optimal segmentation of brain magnetic resonance imaging (MRI). Multilevel thresholding is the most widely used image segmentation method due to its efficacy and simplicity of implementation. The issue, however, is selecting the optimum set of criteria to effectively segment each image. Although methods like Otsu's between-class variance and Kapur's entropy help locate the optimal thresholds, using them for more than two thresholds requires a significant amount of processing resources. Meta-heuristic algorithms are commonly employed in literature to calculate the threshold values; however, they have limitations such as a lack of precise convergence and a tendency to become stuck in local optimum solutions. Due to the aforementioned, we present an improved chameleon swarm algorithm (ICSA) in this paper for image segmentation and global optimization tasks to be able to address these weaknesses. In the ICSA method, the best random mutation strategy has been introduced to improve the searchability of the standard CSA. The best random strategy utilizes three different types of distribution: Levy, Gaussian, and Cauchy to mutate the search individuals. These distributions have different functions, which help enhance the quality of the solutions and avoid premature convergence. Using the IEEE CEC'20 test suite as a recent complex optimization benchmark, a comprehensive set of experiments is carried out in order to evaluate the ICSA method and demonstrate the impact of combining the best random mutation strategy with the original CSA in improving both the performance of the solutions and the rate at which they converge. Furthermore, utilizing the Otsu approach and Kapur's entropy as a fitness function, ICSA is used as an image segmentation method to select the ideal threshold values for segmenting a set of MRI images. Within the experiments, the ICSA findings are compared with well-known metaheuristic algorithms. The comparative findings showed that ICSA performs better than other competitors in solving the CEC'20 test suite and has a significant performance boost in image segmentation. [ABSTRACT FROM AUTHOR]

In this paper, we primarily concentrate on the quasi-projective synchronization of master–slave discrete-time BAM neural networks. Without using the LMI and matrix measure methods, by applying discrete inequalities and the Lyapunov sequences, constructing a discrete inequality group, then solving the inequality group, two novel criteria to guarantee the quasi-projective synchronization are obtained for the considered networks. The controllers designed, the approach, and the results in the paper are fully novel, which advance the development of the study of synchronization of discrete neural networks. [ABSTRACT FROM AUTHOR]

Feature selection (FS) is a crucial preprocessing step that aims to eliminate irrelevant and redundant features, reduce the dimensionality of the feature space, and enhance clustering efficiency and effectiveness. FS is categorized as NP-Hard due to the high number of existing solutions. Various metaheuristic methods have been developed to address the FS problem, yielding promising results. Particularly, particle swarm optimization (PSO), an evolutionary computing (EC) approach guided by swarm intelligence, has gained widespread adoption owing to its implementation simplicity and potential for global search. This paper analyzes several variants of PSO algorithms and introduces a new FS method called HPSO. The proposed approach utilizes an asynchronously adaptive inertia weight and an improved constriction factor. Additionally, it incorporates a chaotic map and a MAD fitness function with a feature count penalty to tackle the clustering FS problem. The efficiency of the developed method is evaluated against the genetic algorithm (GA) and well-known variants of PSO algorithms, including PSOs with fixed inertia weights, PSOs with improved inertia weights, PSOs with fixed constriction factors, PSOs with improved constriction factors, PSOs with adaptive inertia weights, and PSO's includes advanced learning exemplars and sophisticated structure topologies. This paper assesses two different reference text data sets, Reuters-21578 and Webkb. In comparison with competitive methods, the proposed HPSO method achieves higher clustering precision and selects a more informative feature set. [ABSTRACT FROM AUTHOR]

In recent years, the proliferation of vehicles on roadways, accompanied by an escalating demand for augmented safety and refined traffic management, has fueled substantial progress in the domains of computer vision and deep learning methodologies. This technological surge has precipitated ground-breaking applications, with real-time vehicle detection emerging as a linchpin. Capability to accurately detect vehicles in a real-time holds profound implications across diverse domains. In the context of autonomous driving, it underpins essential functions such as path planning, collision avoidance, and decision-making, essential for the safe operation of autonomous vehicles. In the arena of traffic monitoring, it empowers transportation authorities to acquire real-time insights into traffic dynamics, congestion patterns, and incident detection, thereby optimizing traffic flow and resource allocation strategies. Furthermore, within surveillance systems, real-time vehicle detection equips security personnel with the means to swiftly identify potential security threats, enhancing overall situational awareness and safety. This paper contributes to this dynamic landscape by introducing an innovative architecture expressly crafted for vehicle detection. Leveraging state-of-the-art computer vision techniques and deep learning methodologies, this architecture addresses the evolving challenges posed by contemporary traffic scenarios. Through the enhancement of real-time vehicle detection capabilities, our aim is to fortify the foundations of safety, efficiency, and security within vehicular contexts. The paper includes a fusion of AlexNet and Darknet accompanied by a comparative analysis. It is expected to deliver unmatched dependability and accuracy in real-time traffic monitoring, setting a new standard for performance in this application. [ABSTRACT FROM AUTHOR]

Recently, fully transformer-based trackers have achieved impressive tracking results, but this also brings a great deal of computational complexity. Some researchers have applied token pruning techniques to fully transformer-based trackers to diminish the computational complexity, but this leads to missing contextual information that is important for the regression task in the tracker. In response to the above issue, this paper proposes a token fusion method that speeds up inference while avoiding information loss and thus improving the robustness of the tracker. Specifically, the input of the transformer's encoder contains search tokens and exemplar tokens, and the search tokens are divided into tracking object tokens and background tokens according to the similarity between search tokens and exemplar tokens. The tokens with greater similarity to the exemplar tokens are identified as tracking object tokens, and those with smaller similarity to the exemplar tokens are identified as background tokens. The tracking object tokens contain the discriminative features of the tracking object, for the sake of making the tracker pay more attention to the tracking object tokens while reducing the computational effort. All the tracking object tokens are kept, and then, the background tokens are weighted and fused to form new background tokens according to the attention weight of the background tokens to prevent the loss of contextual information. The token fusion method presented in this paper not only provides efficient inference of the tracker but also makes the tracker more robust. Extensive experiments are carried out on popular tracking benchmark datasets to verify the validity of the token fusion method. [ABSTRACT FROM AUTHOR]

With the introduction of ChatGPT, the public's perception of AI-generated content has begun to reshape. Artificial intelligence has significantly reduced the barrier to entry for non-professionals in creative endeavors, enhancing the efficiency of content creation. Recent advancements have seen significant improvements in the quality of symbolic music generation, which is enabled by the use of modern generative algorithms to extract patterns implicit in a piece of music based on rule constraints or a musical corpus. Nevertheless, existing literature reviews tend to present a conventional and conservative perspective on future development trajectories, with a notable absence of thorough benchmarking of generative models. This paper provides a survey and analysis of recent intelligent music generation techniques, outlining their respective characteristics and discussing existing methods for evaluation. Additionally, the paper compares the different characteristics of music generation techniques in the East and West as well as analysing the field's development prospects. [ABSTRACT FROM AUTHOR]

IoT devices convert billions of objects into data-generating entities, enabling them to report status and interact with their surroundings. This data comes in various formats, like structured, semi-structured, or unstructured. In addition, it can be collected in batches or in real time. The problem now is how to benefit from all of this data gathered by sensing and monitoring changes like temperature, light, and position. In this paper, we propose a predictive analytics framework constructed on top of open-source technologies such as Apache Spark and Kafka. The framework focuses on forecasting temperature time series data using traditional and deep learning predictive analytics methods. The analysis and prediction tasks were performed using Autoregressive Integrated Moving Average (ARIMA), Seasonal Autoregressive Integrated Moving Average (SARIMA), Long Short-Term Memory (LSTM), and a novel hybrid model based on Convolution Neural Network (CNN) and LSTM. The purpose of this paper is to determine whether and how recently developed deep learning-based models outperform traditional algorithms in the prediction of time series data. The empirical studies conducted and reported in this paper demonstrate that deep learning-based models, specifically LSTM and CNN-LSTM, exhibit superior performance compared to traditional-based algorithms, ARIMA and SARIMA. More specifically, the average reduction in error rates obtained by LSTM and CNN-LSTM models were substantial when compared to other models indicating the superiority of deep learning. Moreover, the CNN-LSTM-based deep learning model exhibits a higher degree of closeness to the actual values when compared to the LSTM-based model. [ABSTRACT FROM AUTHOR]

Most printed material is produced by printing halftone dot patterns. One of the key issues that determine the attainable print quality is the structure of the paper surface, but the relation is non-deterministic in nature. We examine the halftone print quality and study the statistical dependence between the defects in printed dots and the topography measurement of the unprinted paper. The work concerns SC paper samples printed by an IGT gravure test printer. We have small-scale 2D measurements of the unprinted paper surface topography and the reflectance of the print result. The measurements before and after printing are aligned with subpixel resolution, and individual printed dots are detected. First, the quality of the printed dots is studied using Self Organizing Map and clustering and the properties of the corresponding areas in the unprinted topography are examined. The printed dots are divided into high and low print quality. Features from the unprinted paper surface topography are then used to classify the corresponding paper areas using Support Vector Machine classification. The results show that the topography of the paper can explain some of the print defects. However, there are many other factors that affect the print quality, and the topography alone is not adequate to predict the print quality. [ABSTRACT FROM AUTHOR]

This paper investigates the art design of user interaction experience in mobile systems through the methods of contextual perception and machine learning. The theoretical foundations for the design of intangible cultural heritage interactive display resources include digital display theory, intangible cultural heritage education theory, embodied cognition theory, and gamification design theory. Based on the modelling and analysis of the theoretical foundations, the design principles are derived, including the heritage education principle, the somatic interaction design principle, and the content design principle. In this paper, we use ontologies to construct user knowledge models, fuse multi-situational similarity metrics, screen out candidate neighbour sets through preliminary screening, then combine the user's activity to construct time-based weight tensor scores, use tensor decomposition to obtain recommendation evaluation values, and finally use the recommendation evaluation values to make artistic recommendations. The experimental results show that the algorithm can still obtain a good recommendation implementation in the case of extremely sparse data. The analysis of the fit between augmented reality and folk art appreciation class, as well as the reference to relevant application cases, make design and practice of augmented reality application in folk art appreciation class, try to solve the common problems in folk art appreciation class, analyse the feedback effect and make a summary and outlook. [ABSTRACT FROM AUTHOR]

Microscopic image analysis technology helps solve the inadvertences of artificial traditional methods in disease, wastewater treatment, and environmental change monitoring analysis. Convolutional neural network (CNN) play an important role in microscopic image analysis. Image segmentation, in which U-Net is increasingly applied in microscopic image segmentation, is a crucial step in detection, tracking, monitoring, feature extraction, modelling, and analysis. This paper comprehensively reviews the development history of U-Net, analyses several research results of various segmentation methods since the emergence of U-Net, and conducts a comprehensive review of related papers. This paper summarised the improved methods of U-Net and then listed the existing significance of image segmentation techniques and their improvements introduced over the years. Finally, focusing on the different improvement strategies of U-Net in different papers, the related work of each application target is reviewed according to detailed technical categories to facilitate future research. Researchers can see the dynamics of the transmission of technological development and keep up with future trends in this interdisciplinary field. [ABSTRACT FROM AUTHOR]

In the high-speed railway system, trains' original timetable is often disturbed by some emergencies including geological disasters and equipment failures, which brings great influence to passengers. This paper proposes a real-time high-speed train dispatching model in case of segment blockages, where a railway network is considered. The model includes the following two parts. First, if the trains are not cancelled or decelerated after a blockage occurs, the scope of the affected trains and stations is roughly estimated via the prediction of delay propagation model. Second, with the overall delay as the objective function, this paper constructs a mixed integer nonlinear programming (MINLP) model by considering the following three adjustment strategies: cancellation, delayed departure and deceleration, where the safe headway of the train operation is guaranteed by the moving blocking principle. Furthermore, to reduce the computation complexity, the solution of the model is only considered within the scope obtained in the first stage. The model is verified by using a small railway network with Nanjing as the hub station, which shows that the model is useful for reducing the effect of a disruption on original timetable, especially in comparison with the First Scheduled First Served (FSFS) rule used in practice. [ABSTRACT FROM AUTHOR]

Vision-language navigation (VLN) tasks require an agent to follow language instructions from a human guide to navigate in previously unseen environments using visual observations. This challenging field, involving problems in natural language processing (NLP), computer vision (CV), robotics, etc., has spawned many excellent works focusing on various VLN tasks. This paper provides a comprehensive survey and an insightful taxonomy of these tasks based on the different characteristics of language instructions. Depending on whether navigation instructions are given once or multiple times, we divide the tasks into two categories, i.e., single-turn and multiturn tasks. We subdivide single-turn tasks into goal-oriented and route-oriented tasks based on whether the instructions designate a single goal location or specify a sequence of multiple locations. We subdivide multiturn tasks into interactive and passive tasks based on whether the agent is allowed to ask questions. These tasks require different agent capabilities and entail various model designs. We identify the progress made on these tasks and examine the limitations of the existing VLN models and task settings. Hopefully, a well-designed taxonomy of the task family enables comparisons among different approaches across papers concerning the same tasks and clarifies the advances made in these tasks. Furthermore, we discuss several open issues in this field and some promising directions for future research, including the incorporation of knowledge into VLN models and transferring them to the real physical world. [ABSTRACT FROM AUTHOR]

One of the most widely referenced Swarm Intelligence (SI) algorithms is the Grey Wolf Optimizer (GWO), which is based on the pack hunting and natural leadership organization of grey wolves. The GWO algorithm offers several significant benefits, including simple implementation, rapid convergence, and superior convergence outcomes, leading to its effective application in diverse fields for solving optimization issues. Consequently, the GWO has rapidly garnered substantial research interest and a broad audience across numerous areas. To better understand the literature on this algorithm, this review paper aims to consolidate and summarize research publications that utilized the GWO. The paper begins with a concise introduction to the GWO, providing insight into its natural establishment and conceptual framework for optimization. It then lays out the theoretical foundation and key procedures involved in the GWO, following which it comprehensively examines the most recent iterations of the algorithm and categorizes them into parallel, modified, and hybridized variations. Subsequently, the primary applications of the GWO are thoroughly explored, spanning various fields such as computer science, engineering, energy, physics and astronomy, materials science, environmental science, and chemical engineering, among others. This review paper concludes by summarizing the key arguments in favour of GWO and outlining potential lines of inquiry in the future research. [ABSTRACT FROM AUTHOR]

The coati optimization algorithm (COA) is a recently proposed heuristic algorithm. The COA algorithm, which solved the continuous optimization problems in its original paper, has been converted to a binary optimization solution by using transfer functions in this paper. Thus, binary COA (BinCOA) is proposed for the first time in this study. In this study, twenty transfer functions are used (four S-shaped, four V-shaped, four Z-shaped, four U-shaped, and four taper-shaped transfer functions). Thus, twenty variations of BinCOA are obtained, and the effect of each transfer function on BinCOA is examined in detail. The knapsack problem (KP) and uncapacitated facility location problem (UFLP), which are popular binary optimization problems in the literature, are chosen to test the success of BinCOA. In this study, small-, middle-, and large-scale KP and UFLP datasets are selected. Real-world problems are not always low-dimensional. Although a binary algorithm sometimes shows superior success in low dimensions, it cannot maintain the same success in large dimensions. Therefore, the success of BinCOA has been tested and demonstrated not only in low-dimensional binary optimization problems, but also in large-scale optimization problems. The most successful transfer function is T3 for KPs and T20 for UFLPs. This showed that S-shaped and taper-shaped transfer functions obtained better results than others. After determining the most successful transfer function for each problem, the enhanced BinCOA (EBinCOA) is proposed to increase the success of BinCOA. Two methods are used in the development of BinCOA. These are the repair method and the XOR gate method. The repair method repairs unsuitable solutions in the population in a way that competes with other solutions. The XOR gate is one of the most preferred methods in the literature when producing binary solutions and supports diversity. In tests, EBinCOA has achieved better results than BinCOA. The added methods have proven successful on BinCOA. In recent years, the newly proposed evolutionary mating algorithm, fire hawk optimizer, honey badger algorithm, mountain gazelle optimizer, and aquila optimizer have been converted to binary using the most successful transfer function selected for KP and UFLP. BinCOA and EBinCOA have been compared with these binary heuristic algorithms and literature. In this way, their success has been demonstrated. According to the results, it has been seen that EBinCOA is a successful and preferable algorithm for binary optimization problems. [ABSTRACT FROM AUTHOR]

This paper aims to solve the Chinese Postman Problem (CPP) using an Ant Colony Optimization (ACO) algorithm. In graph theory, the CPP looks for the shortest closed path that visits every edge of a connected undirected graph. This problem has many applications, including route optimization, interactive system analysis, and flow design. Although numerous algorithms aimed at solving CPP are present in the literature, very few meta-heuristic algorithms are proposed, and no ACO applications have been proposed to solve them. This paper tries to fill this gap by presenting an ACO algorithm that solves CPP (ACO-CPP). To prove its consistency and effectiveness, ACO-CPP is compared with a Genetic Algorithm (GA) and a recursive algorithm throughout three experiments: (1) recursive-ACO-GA comparisons over randomly generated graphs for the attainment of the global optimum; (2) ACO-GA statistical comparisons over specifically generated graphs; (3) recursive-ACO-GA comparisons by changing ACO hyperparameters over randomly generated graphs for the attainment of the global optimum. The experiments prove that the ACO-CPP algorithm is efficient and exhibits a consistency similar to GA when the number of possible solutions to explore is relatively low. However, when that number greatly exceeds those explored, ACO outperforms GA. This suggests that ACO is more suitable for solving problems with a CPP structure. [ABSTRACT FROM AUTHOR]

Object detection is one of the core tasks of computer vision, and bounding box (bbox) regression is one of the basic tasks of object detection. In recent years of related research, bbox regression is often used in the Intersection over Union (IoU) loss and its improved version. In this paper, for the first time, we introduce the Dice coefficient into the regression loss calculation and propose a new measure which is superior to and can replace the IoU. We define three properties of the new measure and prove the theory by mathematical reasoning and analysis of the existing work. This paper also proposes the N-IoU regression loss family. And the superiority of the N-IoU regression loss family is proved by designing simulation experiments and comparative experiments. The main results of this paper are: (1) The proposed new measure is better than IoU which can be used to evaluate bounding box regression, and the three properties of the new measure can be used as a broad criterion for the design of regression loss functions; and (2) we propose N-IoU loss. The parameter n of N-IOU can be debugged, which can be widely adapted to different application scenarios with higher flexibility, and the regression performance is better. [ABSTRACT FROM AUTHOR]

This paper presents an intelligent healthcare framework by incorporating modern computing technologies like machine learning and deep learning. The sole motivation of this paper is to predict diabetes status intelligently so that patients can be made aware of their diabetes status regularly. A novel ensemble machine learning-based diabetes prediction framework, namely "iDP", has been proposed in this paper. Six different machine learning approaches have been used for the iDP framework, namely random forest, decision tree, neural network, AdaBoost, support vector machine, and XGBoost. Training and testing of ensemble modeling have been performed based on these learning techniques. A comprehensive preliminary data analysis and screening of statistical analysis have been systematically demonstrated in this paper. The recursive feature elimination and fivefold cross-validation methods have been used to select the best model. All experiments have been performed using R-Studio 1.2.5 with the help of the Rattle library 5.4.0. The preliminary data analytic and Pearson correlation coefficient values have been computed to verify the linearity relationship among the data. Results have been reported in eight performance metrics, including accuracy, specificity, sensitivity, area under curve, area under convex hull, minimum error rate, minimum weighted loss, and precision. The proposed iDP framework improves the results on different aspects in order to compare them with individual learning approaches. Despite this, a comparative performance analysis of the iDP framework has also been performed with state-of-the-art prevailing models. As per experiment, the iDP framework gives better results in most cases, like accuracy: 95.26%, AUC: 91.15%, sensitivity: 96.81%, specificity: 97.72% and precision: 90.72%. [ABSTRACT FROM AUTHOR]

The emergence of digital media has changed the way people live and learn. In the context of the new era, digital media is gradually integrating into people's life and learning. Digital media contains massive images, and fine-grained image recognition for digital media has become an important topic. The challenge of fine-grained image recognition is that the difference between different categories is small, and the difference between the same categories is sometimes large. This work designs a fine-grained image recognition based on feature enhancement (FIRFE). This extracts as much information as possible from fine-grained images under weak supervision to improve the recognition accuracy. When the existing methods extract image features, the feature extraction other than the most significant local feature is not enough. This deals with local features alone and ignores the relationship between features. First, this paper designs a feature enhancement and suppression module to process image features. Secondly, this paper designs pyramid residual convolution. This uses different scale convolution kernels to capture different levels of features in the scene. Thirdly, this paper uses the softpool method to rationally allocate the information weight in the pooling process. Fourth, this paper uses feature focus module to mine more features. This focuses on obtaining similar information in multiple local features as discriminant features to further improve the recognition. Fifthly, this paper carried out systematic experiments on the designed method. The proposed method achieves 94.3%/95.7% accuracy, 92.9%/94.1% recall, and 91.4%/92.2% F1 score on different datasets. This verified the superiority of this method for fine-grained image recognition of digital media. [ABSTRACT FROM AUTHOR]

Manual term extraction is similar to literal meaning: A translator browses text, classifies words, and prepares for translation. Terminology, as a centralized carrier of expertise, creation, popularization, and disappearance, dynamically reflects the development and evolution of an industry. The automatic extraction of terminology is a key technology for creating a professional terminology database, and it is also a key topic in the field of natural language processing. The purpose of this paper is to study how to analyse a term extraction algorithm based on machine learning and a comprehensive feature strategy. Focusing on the problems of poor generality and single statistical features of current term extraction algorithms, this paper proposes an improved domain ontology term extraction algorithm based on a comprehensive feature strategy. Moreover, automatic term extraction experiments based on a word-based maximum entropy model and a conditional random field model based on machine learning are conducted in this paper. Its word-based conditional random field model outperforms the maximum entropy model. The experimental results show that the algorithm based on the comprehensive feature strategy improves the accuracy by 8.6% compared with the TF-IDF algorithm and the C-value term extraction algorithm. This algorithm can be used to effectively extract the terms in a text and has good generality. [ABSTRACT FROM AUTHOR]

With the rapid development of information technology in modern society, the application of multimedia integration platform is more and more extensive. Speech recognition has become an important subject in the process of multimedia visual interaction. The accuracy of speech recognition is dependent on a number of elements, two of which are the acoustic characteristics of speech and the speech recognition model. Speech data is complex and changeable. Most methods only extract a single type of feature of the signal to represent the speech signal. This single feature cannot express the hidden information. And, the excellent speech recognition model can also better learn the characteristic speech information to improve performance. This work proposes a new method for speech recognition in multimedia visual interaction. First of all, this work considers the problem that a single feature cannot fully represent complex speech information. This paper proposes three kinds of feature fusion structures to extract speech information from different angles. This extracts three different fusion features based on the low-level features and higher-level sparse representation. Secondly, this work relies on the strong learning ability of neural network and the weight distribution mechanism of attention model. In this paper, the fusion feature is combined with the bidirectional long and short memory network with attention. The extracted fusion features contain more speech information with strong discrimination. When the weight increases, it can further improve the influence of features on the predicted value and improve the performance. Finally, this paper has carried out systematic experiments on the proposed method, and the results verify the feasibility. [ABSTRACT FROM AUTHOR]

Sub-health state is a state of health and low quality between disease and health. The theoretical basis of children's sub-health is to start from the whole. The common clinical sub-health conditions cannot be explained by modern detection methods, and it can be screened and analyzed with the help of big data in medical informatization. The combination of "Internet + " and the health care model is an innovation in the construction of medical informatization. It can provide many considerate services to the masses in time and alleviate the anxiety of illness. Therefore, it is very necessary to carry out the efficacy evaluation of children's sub-health from the perspective of medical information. Therefore, this paper completes the following work with the help of AI neural network: (1) This paper proposes an improved AlexNet network evaluation method based on attention mechanism. In this study, attention mechanism is added to the original AlexNet model to weight each channel of the feature layer. At the same time, we improve the large convolution kernel of the previous layers of the original AlexNet network and use batch normalization instead of the local response normalization (LRN) layer in the original model. (2) This paper proposes an evaluation method based on improved residual network, which improves the original residual block of the residual model and widens the residual block. The residual block can effectively reduce the amount of network parameters and improve the efficiency of network training. (3) This paper proposes an evaluation method of multi-scale feature fusion (MSFF). The features extracted from the improved AlexNet and residual network are fused and then evaluated. At this time, the training time is greatly shortened, and the accuracy is higher than that of single model. [ABSTRACT FROM AUTHOR]

This paper explores the knowledge network structure of foreign research literature by applying the qualitative comparative analysis (QCA) method to the field of information science and library science (ISLS) from the perspective of the cocitation of social network actors such as authors, institutions, countries, and literature, and it further reveals the future application trends of this method. [Method/process] Based on 86 journals in the ISLS field that were downloaded from the Web of Science using the QCA method, the social network analysis (SNA) method and the visual analysis tool Gephi are used to analyse the author cooperation network, the research institution cooperation network, the national cooperation network, the cocitation network, the cutting-edge trends, etc., of journal papers. The analysis shows that the QCA method covers a wide range within the field of ISLS, but the research topics involved in this field are not concentrated, and the author cooperation network has scale-free characteristics. The application of the QCA method is still dominant in European and American countries, and China, the USA, and Italy all play key roles in the national cooperation network. Finally, the institutional cooperation network has certain small group attributes. [ABSTRACT FROM AUTHOR]

With the ongoing integration of information technology and industrialization, strategic emerging industries are becoming an increasingly important force in guiding future economic and social development. As one of the strategic emerging industries' development priorities and a replacement for scarce labor resources, industrial robots will be widely used in labor-intensive industries. It is a contentious topic how to evaluate and improve robot knowledge education. There are modular features introduced in this paper, which builds a modular network for the evaluation of robot knowledge education quality and enhances the cognitive computing ability of artificial neural networks and their ability to process complex information. A modular neural network-based model of feature combination robot knowledge education quality evaluation is developed based on the three aspects of module division method, subnet structure selection, and feature combination output. The following are some of the paper's most important contributions: K-OD algorithm of density clustering optimized by K-means is proposed. Because of its high level of modularized partition simulating, this method has an excellent clustering effect, and it identifies the core points, boundary points, as well as outliers. Using K-OD algorithm, the calculation of density radius and threshold is optimized by using density clustering, which reduces the overall computational complexity. Find out how SOM neural networks learn from competition. SOM network's competition layer neuron weights are prone to falling into local optimal solutions, so an SASOM neural network with weight adjustment simulated by an annealing algorithm is proposed to address this issue. It is more accurate in terms of prediction and error, and it is better at identifying sample attribute features. This work builds a modular neural network for evaluating robot knowledge education quality using K-OD clustering algorithm and SASOM neural network, which introduces the simulated annealing mechanism. [ABSTRACT FROM AUTHOR]

The transmission line is one of the most crucial electric power system components, demanding special attention since they are subject to failures that can cause disruptions in energy supply. In this scenario, the fault location emerges as a fundamental task, providing an approximate position where the failure occurred in the line. This paper presents a method for fault location using a novel dynamic regressor selection (DRS) framework, in which we introduce embedded normalization, incorporating the data scaling process inside the framework. The DRS aims to select the most accurate models from a pool of regressors to predict the fault distance. Moreover, since there is a lack of a public dataset, this paper presents and makes a new fault dataset available to the scientific community with several features extracted from current and voltage signals of representative failure events. In our experiments, we demonstrate the importance of this embedded normalization as well as the significance in the variation of critical hyperparameters of the DRS strategy, such as the distance metric used to define the region of competence and the criterion to select the best regressors from the pool of predictors. This work also presents the definition of the oracle concept in DRS, which represents the ideal predictor selection scheme. The results demonstrate the effectiveness of the proposed method with a mean error of 0.7086 km ± 0.0068 km, representing 0.1712% ± 0.0016% of the transmission line extension (414 km). [ABSTRACT FROM AUTHOR]

This letter identifies a fundamental mistake in the published paper titled "A projective general linear group-based algorithm for the construction of substitution box for block ciphers." The projective general linear group-based algorithm is designed to provide high nonlinearity, low differential uniformity, and good resistance against linear and differential attacks. The projective general linear group-based algorithm uses the linear fractional transformation defined over finite field GF 2 8 . The authors of the paper claimed that the irreducible polynomial of degree 8 is used for the construction of a finite field GF 2 8 . We have observed that the projective general linear group-based algorithm uses a reducible polynomial of degree 8. Thus, the inverse of each element does not exist in the finite field GF 2 8 . We have also shown that the values of the S-Box given in the paper are also incorrect. [ABSTRACT FROM AUTHOR]

The adaptive predefined settling time performance tracking control for a class of uncertain nonlinear network systems with fast time-varying or even abrupt reference signals is studied. In this paper, an improved performance function with time-varying boundary (following the change of reference signal) is designed to overcome the problem that the traditional prescribed performance control (PPC) can only converge to a constant in steady state. Compared with the existing adaptive finite time prescribed performance congestion control for tracking fixed value reference signals, an adaptive predefined settling time prescribed performance congestion control strategy is proposed by means of some new recursive construction (introducing nonautonomous differential equations (NDE)) and analysis innovation. In addition, the network system considered in this paper is more general. Under the framework of backstepping, NDE is used iteratively to control the approximation error with boundary inequality. It is proved theoretically that all signals of the closed-loop system are predefined settling time bounded. Finally, the effectiveness of the proposed control strategy is verified by simulations. [ABSTRACT FROM AUTHOR]

A continuous monitoring of neurological status can help in reporting the physical and mental health of a person. This can be capitalized for building a healthcare tracking system using a wearable device and a handheld mobile device. In this paper, we have used the non-EEG physiological biosignals dataset which gives practicability among subjects for acquiring data easily from wearable device sensors linearly and comfortably rather than the way of putting the subjects in a cumbersome setup laboratory. This paper proposes a custom fully convolutional-LSTM (FCN-LSTM) network to identify the neurological status of a subject using multivariate time series physiological sensor data. The proposed architecture uses parallel stacks of the convolutional layers and LSTM cells. This combination of different network types is significant for the selected problem as the fully convolutional section of the model extracts the local spatial features in the data, while the LSTM network handles the high-level features and temporal dependencies. The proposed FCN-LSTM model yielded a high accuracy of 98.6% and a precision of 98% on the non-EEG dataset from UT-Dallas. The average accuracy of single-subject results of the dataset using the proposed model was observed to be 99.26%. The results from the proposed model are significantly improved when compared with various state-of-the-art works on this problem. These results strongly suggest that this model, when put on a wearable device, can be effectively used to detect the neurological status or stress that the subject may be going through in real time. [ABSTRACT FROM AUTHOR]

Deep learning has proven to be highly effective for human activity recognition (HAR) when large amount of labelled data is available for the target task. However, training a deep learning model to generalize well on a new task with just-few observations of labelled data is an active area of research. In this paper, a novel few-shot transfer learning (FSTL) approach is proposed for classification of human activities using just few instances (shots) of the data obtained from a wearable system assembled to collect inertial sensor data for different human activities, performed by two users. First, a deep learning model is trained on a large publicly available HAR dataset. The model parameters of such a model are then fine-tuned using the Reptile algorithm to determine the optimal initial parameter set using which, the model will classify activities with just few-shots of data from the target task. The proposed FSTL approach yields an average classification accuracy of 74.86 ± 0.71% and 79.20 ± 1.05% for 3-way, 5-shot classification of new activities performed by a single user and same set of activities performed by a new user, respectively. When the pre-trained weights are used as the initial weights in the Reptile algorithm, the generalization ability of the model improves by about 10% for 3-way, 5-shot classification as compared to using few-shot learning without parameter transfer. [ABSTRACT FROM AUTHOR]

Multimodal sentiment analysis (MSA) seeks to understand human affection by leveraging signals from multiple modalities. A core challenge in MSA is the effective extraction of sentimental relations between these signals, as this can enhance a model's consistency and accuracy. Existing studies typically use multimodal matching tasks to learn all semantic relations between modalities and then use downstream task to obtain the specific semantics from the multimodal representation. However, there are multiple semantics between modalities, such as action semantics, scene semantics and sentiment semantics. Relying solely on specific tasks to filter these semantics often results in a surplus of redundant information in the multimodal representation, potentially degrading MSA accuracy. In addition, the unimodal semantic expression is also important. In this paper, we propose a semantic-specific multimodal relation learning method to correlate modalities with specific semantics. Specifically, with smaller computational resources, we enhance unimodal sentimental semantic expression while diminishing non-sentimental semantic information in the multimodal representation. We conducted experiments on multimodal sentiment analysis datasets, CMU-MOSI, CMU-MOSEI and CH-SIMS. The results show that our method outperforms the current state-of-the-art. Notably, on the Acc2 evaluation metric, our approach exhibits an average accuracy improvement of 0.75 compared to the best baseline. [ABSTRACT FROM AUTHOR]

Visual XAI methods enable experts to reveal importance maps highlighting intended classes over input images. This research paper presents a novel approach to visual explainable artificial intelligence (XAI) for object detection in deep learning models. The study investigates the effectiveness of activation maps generated by five different methods, namely GradCAM, GradCAM++, EigenCAM, HayCAM, and a newly proposed method called "HayCAMJ", in detecting objects within images. The experiments were conducted on two datasets (Pascal VOC 2007 and Pascal VOC 2012) and three models (ResNet18, ResNet34, and MobileNet). Zero padding was applied to resize and center the objects due to the large objects in the images. The results show that HayCAMJ performs better than other XAI techniques in detecting small objects. This finding suggests that HayCAMJ has the potential to become a promising new approach for object detection in deep classification models. [ABSTRACT FROM AUTHOR]

QT interval carries essential information about ventricular depolarization and repolarization. Therefore, in many investigations, it is vital to determine the QT interval accurately and monitor the corresponding QT interval variations. This paper presents a novel implementation of the Generative Adversarial Networks (GAN) to predict the QT waveform interval from the electrocardiogram. The network then estimates the Q and T locations to determine the QT interval. The accuracy of the proposed method in measuring the QT interval was 94.10%. The procedure is evaluated against ground truth performed by expert physicians, and the limits of agreement are used as the measure of performance. The bias in the estimate of the Q locations was 2.92 ms, with 11.48 ms as an upper limit of agreement and − 5.65 ms as the lower limit. The bias in estimating the T location was 1.14 ms with an upper limit of 15.04 ms and a lower limit of − 12.79 ms. Moreover, the bias in the estimate of the QT interval was − 7.10 ms with an upper limit of 50.17 ms and a lower limit of − 64.38 ms. These limits of agreements are clinically acceptable and show that the estimate of the QT interval agrees with the expert's annotations. [ABSTRACT FROM AUTHOR]

Vision-based Human Action Recognition is difficult owing to the variations in the same action performed by various people, the temporal variations in actions, and the difference in viewing angles. Researchers have recently adopted multi-modal visual data fusion strategies to address the limitations of single-modality methodologies. Many researchers strive to produce more discriminative features because most existing techniques' success relies on feature representation in the data modality under consideration. Human action consists of several sub-actions whose duration vary between individuals. This paper proposes a multifarious learning framework employing action data in depth and skeleton formats. Firstly, a novel action representation named Multiple Sub-action Enhanced Depth Motion Map (MS-EDMM), integrating depth features from overlapping sub-actions, is proposed. Secondly, an efficient method is introduced for extracting spatio-temporal features from skeleton data. This is achieved by dividing the skeleton sequence into sub-actions and summarizing skeleton joint information for five distinct human body regions. Next, a multi-stream deep learning model with Attention-guided CNN and residual LSTM is proposed for classification, followed by several score fusion operations to reap the benefits of streams trained with multiple data types. The proposed method demonstrated a superior performance of 1.62% over an existing method that utilized skeleton and depth data, achieving an accuracy 89.76% on a single-view UTD-MHAD dataset. Furthermore, on the multi-view NTU RGB+D dataset demonstrated encouraging performance with an accuracy of 89.75% in cross-view and 83.8% in cross-subject evaluations. [ABSTRACT FROM AUTHOR]

Brain–Computer Interface is tested as a successful method in improving human cognitive functions such as attention and memory. Attention plays a significant role in areas ranging from a person's day-to-day life to educational domain and professional activities. When attention is evaluated using camera-based techniques, the users may suffer privacy issues. Using Brain–Computer Interface (BCI) to obtain a measure of attention will be useful in this regard. The paper proposes a Human Attention Recognition System (HARS) in which EEG signal acquisition is used to obtain the attention of the individual, Renyi's entropy-based mutual information method is used for feature selection and a deep learning-based classifier is used to classify the signals. HADS is not camera-based: therefore, faces of the subjects are not revealed. EEG signals were collected using the Ultracortex Mark III dry electrodes and were visualised using OpenBCI GUI (Graphical User Interface). The experiment is validated using the publicly available Confused Student EEG dataset from Kaggle, giving an accuracy of 99.21%. The results indicate that the proposed method can identify attention levels accurately and can be effectively used in educational systems, biofeedback and medical research. [ABSTRACT FROM AUTHOR]

Point clouds in the real world are often sparse and incomplete. Point cloud completion aims to restore incomplete point clouds into meaningful shapes. In recent years, point cloud completion has attracted the interest of many researchers. In most existing methods, the efficiency of completion is generally ignored in favor of producing meaningful shapes with adequate details. This paper proposes a fast point completion network (FPCN). FPCN is mainly composed of a multi-scale attention encoder (MSAE) and a structural refinement (SR) module. MSAE first obtains multi-scale geometric information by extracting incomplete inputs at different resolutions. After that, MSAE fuses multi-scale geometric information through cross-attention mechanisms. Compared with existing mainstream encoders, MSAE can extract rich geometric information from input point clouds with low complexity. The SR module aims to extract local information from the coarse point clouds to guide the process of extending points. Furthermore, the process of extending points is achieved by a replication strategy. Compared to existing folding-based decoders, the SR module can produce fine point clouds with more local details. Compared to existing transformer-based decoders, the SR module has a lower calculation price by employing a replication strategy to generate high-resolution point clouds. In conclusion, FPCN can restore partial point clouds into meaningful shapes efficiently, and the outcomes of completion contain sufficient details. Extensive experiments on various datasets demonstrate the performance and efficiency of the FPCN in point cloud completion. Source code is available at https://github.com/doldolOuO/FPCN. [ABSTRACT FROM AUTHOR]

The increasing reliance on internet-based services has heightened the vulnerability of network infrastructure to cyberattacks, particularly distributed denial of service (DDoS) attacks. These attacks can cause severe disruptions and significant financial losses. Early detection of malicious traffic is crucial in effectively combating such threats. This paper presents an innovative approach called the Encoder-Stacked deep neural networks (E-SDNN) model, which leverages Stacked/bagged multi-layer perceptrons (MLP) for accurate DDoS attack detection. The proposed method employs an encoder to select pertinent features from a preprocessed dataset, enabling precise attack detection. Extensive experiments were conducted on benchmark cybersecurity datasets, namely CICDS2017 and CICDDoS2019, encompassing various DDoS attack scenarios. The experimental results demonstrate the superiority of the E-SDNN model compared to state-of-the-art methods. The proposed E-SDNN model achieved an impressive overall accuracy rate of 99.94% and 98.86% for CICDDS2017 and CICDDoS2019, respectively. [ABSTRACT FROM AUTHOR]

Limitations in computer-assisted diagnosis include lack of labeled data and inability to model the relation between what experts see and what computers learn. Even though artificial intelligence and machine learning have demonstrated remarkable performances in medical image computing, their accountability and transparency level must be improved to transfer this success into clinical practice. The reliability of machine learning decisions must be explained and interpreted, especially for supporting the medical diagnosis. While deep learning techniques are broad so that unseen information might help learn patterns of interest, human insights to describe objects of interest help in decision-making. This paper proposes a novel approach, DeepCraftFuse, to address the challenge of combining information provided by deep networks with visual-based features to significantly enhance the correct identification of cancerous tissues in patients affected with Barrett's esophagus (BE). We demonstrate that DeepCraftFuse outperforms state-of-the-art techniques on private and public datasets, reaching results of around 95% when distinguishing patients affected by BE that is either positive or negative to esophageal cancer. [ABSTRACT FROM AUTHOR]

This paper presents a novel hybrid optimization method to solve the resource allocation problem for multi-target multi-sensor tracking of drones. This hybrid approach, the Improved Prairie Dog Optimization Algorithm (IPDOA) with the Genetic Algorithm (GA), utilizes the strengths of both algorithms to improve the overall optimization performance. The goal is to select a set of sensors based on norms of weighted distances cost function. The norms are the Euclidean distance and the Mahalanobis distance between the drone location and the sensors. The second one depends on the predicted covariance of the tracker. The Extended Kalman Filter (EKF) is used for state estimation with proper clutter and detection models. Since we use Multi-objects to track, the Joint Probability Distribution Function (JPDA) estimates the best measurement values with a preset gating threshold. The goal is to find a sensor or minimum set of sensors that would be enough to generate high-quality tracking based on optimum resource allocation. In the experimentation simulated with Stone Soup, one radar among five radars is selected at every time step of 50-time steps for 200 tracks distributed over 20 different ground truths. The proposed IPDOA provided optimum solutions for this complex problem. The obtained solution is an optimum offline solution that is used to select one or more sensors for any future flights within the vicinity of the 5 radars. Environment and conditions are assumed to be similar in future drone flights within the radars' defined zone. The IPDOA performance was compared with the other 8 metaheuristic optimization algorithms and the testing showed its superiority over those techniques for solving this complex problem. The proposed simulated model can find the most relevant sensor(s) capable of generating the best quality tracks based on weighted distance criteria (Euclidean and Mahalanobis). That would cut down the cost of operating extra sensors and then it would be possible to move them to other vicinity. [ABSTRACT FROM AUTHOR]

This paper proposes a new method to improve the training efficiency of deep convolutional neural networks. During training, the method evaluates scores to measure how much each layer's parameters change and whether the layer will continue learning or not. Based on these scores, the network is scaled down such that the number of parameters to be learned is reduced, yielding a speed-up in training. Unlike state-of-the-art methods that try to compress the network to be used in the inference phase or to limit the number of operations performed in the back-propagation phase, the proposed method is novel in that it focuses on reducing the number of operations performed by the network in the forward propagation during training. The proposed training strategy has been validated on two widely used architecture families: VGG and ResNet. Experiments on MNIST, CIFAR-10 and Imagenette show that, with the proposed method, the training time of the models is more than halved without significantly impacting accuracy. The FLOPs reduction in the forward propagation during training ranges from 17.83% for VGG-11 to 83.74% for ResNet-152. As for the accuracy, the impact depends on the depth of the model and the decrease is between 0.26% and 2.38% for VGGs and between 0.4 and 3.2% for ResNets. These results demonstrate the effectiveness of the proposed technique in speeding up learning of CNNs. The technique will be especially useful in applications where fine-tuning or online training of convolutional models is required, for instance because data arrive sequentially. [ABSTRACT FROM AUTHOR]

Ensuring the safety of workers and machinery during operations is a critical task in the construction sites. However, an inevitable circumstance in construction sites is the complex and dynamic environment, which often leads to occlusions. When detecting occluded objects in construction sites, general vision-based approaches tend to exhibit lower accuracy and may even miss detections, resulting in potential safety hazards. To handle this issue, this paper proposes a vision-based approach for detecting occluded objects in construction sites. Firstly, the proposed detection algorithm adopts the state-of-the-art YOLOv7 as its backbone. To enhance its capability in capturing contextual information of occluded objects, a novel channel attention mechanism is employed. Then, a design scheme for the detector head is provided by integrating a novel loss function Scylla-Intersection over Union (SIoU) and the non-maximum suppression (NMS) strategy. With the help of the loss function SIoU, the network can compute the loss values of occluded objects more accurately. To ensure that the network can select the right predicted box which closely aligns with the ground truth, the Euclidean distance is utilized as spatial penalty factor during the NMS stage. By implementing these two strategies, the proposed method can preserve both the category information and bounding boxes of occluded objects, which makes them possible to be detected. Finally, detailed experiments are done to verify the proposed method. Experimental results demonstrate that the proposed method has the potential for improving the detection accuracy. Moreover, it shows a better performance in detecting occluded objects in the dynamic construction sites compared to the existing baselines. [ABSTRACT FROM AUTHOR]

Multi-Object Tracking, also known as Multi-Target Tracking, is an important area of computer vision with various applications in different domains. The advent of deep learning has had a profound impact on this field, forcing researchers to explore innovative avenues. Deep learning methods have become the cornerstone of today's state-of-the-art solutions, consistently delivering exceptional tracking results. However, the significant computational demands of deep learning models require powerful hardware resources that do not always match real-time tracking requirements, limiting their practical applicability in real-world scenarios. Thus, there is an imperative to strike a balance by merging robust deep learning strategies with conventional approaches to enable more accessible, cost-effective solutions that meet real-time requirements. This paper embarks on this endeavor by presenting a hybrid strategy for real-time multi-target tracking. It effectively combines a classical optical flow algorithm with a deep learning architecture tailored for human crowd tracking systems. This hybrid approach achieves a commendable balance between tracking accuracy and computational efficiency. The proposed architecture, subjected to extensive experimentation in various settings, demonstrated notable results, achieving a Mean Object Tracking Accuracy (MOTA) of 0.608. This level of performance placed it as the highest ranking solution on the MOT15 benchmark, surpassing the state-of-the-art benchmark of 0.549, and consistently ranked among the superior models on the MOT17 and MOT20 benchmarks. Additionally, the incorporation of the optical flow phase resulted in a substantial reduction in processing time, nearly halving the duration, while simultaneously maintaining accuracy levels comparable to established techniques. [ABSTRACT FROM AUTHOR]

The data-driven modular deep fuzzy model has demonstrated excellent forecasting performance due to its clear architecture and powerful fuzzy inference ability. However, the fixed structure predesigned for specific types of datasets limits the further improvement of generalization ability, especially when facing high-dimensional datasets with the challenging problem of rule explosion, thereby suffering from the time-consuming training process. To overcome these limitations, a novel randomized fuzzy modules (RFMs) stacked residual deep fuzzy system (RFM-RDFS) is proposed in this study. Firstly, to obtain high-precision outputs, this paper presents a residual deep fuzzy framework by stacking RFMs layer by layer, where the modules in the current layer only learn the residual of the previous layer. Secondly, to improve the design freedom for different applications, the RFM with multiple inputs is raised, where the input variables are allocated by a sliding window along the input spaces of the layers, so the number of the parallel modules in the same layer can be flexibly adjusted. Meanwhile, in the RFM, inspired by the extreme learning machine, the centers and widths of Gaussian membership functions are randomly generated, which can greatly accelerate the training speed. Thirdly, to strengthen the approximation performance, the consequent parameters of the fuzzy rules in each RFM are learned by minimizing the loss function with the regularization terms through the ridge regression algorithm. Finally, extensive experiments on various datasets are conducted to compare the proposed RFM-RDFS with several state-of-the-art shallow and deep models. The detailed comparison results clearly show that the proposed RFM-RDFS is more flexible and compact, and has greater advantages in terms of the interpretability, generalization ability and robustness. [ABSTRACT FROM AUTHOR]

Over the preceding decade, there has been a discernible surge in the prominence of artificial intelligence, marked by the development of various methodologies, among which deep learning emerges as a particularly auspicious technique. The captivating attribute of deep learning, characterised by its capacity to glean intricate feature representations from data, has served as a catalyst for pioneering approaches and methodologies spanning a multitude of domains. In the face of the burgeoning exponential growth in digital medical image data, the exigency for adept image compression methodologies has become increasingly pronounced. These methodologies are designed to preserve bandwidth and storage resources, thereby ensuring the seamless and efficient transmission of data within medical applications. The critical nature of medical image compression accentuates the imperative to confront the challenges precipitated by the escalating deluge of medical image data. This review paper undertakes a comprehensive examination of medical image compression, with a predominant focus on sophisticated, research-driven deep learning techniques. It delves into a spectrum of approaches, encompassing the amalgamation of deep learning with conventional compression algorithms and the application of deep learning to enhance compression quality. Additionally, the review endeavours to explicate these fundamental concepts, elucidating their inherent characteristics, merits, and limitations. [ABSTRACT FROM AUTHOR]

With the rapid development of data acquisition and storage technology, spatio-temporal (ST) data in various fields are growing explosively, so many ST prediction methods have emerged. The review presented in this paper mainly studies the prediction of ST series. We propose a new taxonomy organized along three dimensions: ST series prediction methods (focusing on time feature learning, focusing on spatial feature learning, and focusing on spatial–temporal feature learning), techniques of ST series prediction (the RNN-, CNN-, and transformer-based models, as well as the CNN-based-composite model and GNN-based-composite models, and the miscellaneous model) and ST series prediction results (single target and multi-target). We first introduce and explain each dimension of the taxonomy in detail. After providing this three-dimensional view, we comprehensively review and compare the recent related ideas in the literature and analyze their advantages and limitations. Moreover, we summarize the key information of the existing literature and provide guidance for researchers to select suitable models. Second, we summarize the different applications of deep learning models in ST series prediction based on current literature and list relevant datasets and download links per application classifications. Lastly, we comprehensively analyze the current innovation and challenges and suggest future directions for researching ST series prediction after comparing and analyzing the computing performance of these forecasting models. In addition, each method or model solves one aspect of the challenge, which means that two or more methods should be combined to solve more challenges at the same time. We hope this article provides readers a broader and deeper understanding of the field of ST series research. [ABSTRACT FROM AUTHOR]

Fault detection plays a crucial role in ensuring the safety, availability, and reliability of modern industrial processes. This study focuses on data-driven fault detection methods, which have gained significant attention across various industrial sectors due to the rapid development of industrial automation technologies and the availability of extensive datasets. The objectives of this paper are to comprehensively review and present the theoretical foundations of widely used data-driven fault detection approaches. Specifically, these approaches are applied to fault detection in wind turbine systems, with performance evaluation conducted using multiple statistical measures. The data utilized in this study were collected from a simulated benchmark of a wind turbine system. The data-driven methods are tested under the assumption that the wind turbine operates in a steady-state region. Additionally, a comparative study is conducted to identify and discuss the primary challenges associated with the practical application of these methods in real-world scenarios. Simulation results show the effectiveness and efficacy of data-driven approaches concerning the sensitivity and robustness of wind turbine sensor faults as applied in practical industrial environments. [ABSTRACT FROM AUTHOR]

This paper presents a method to perform the real-time creation of models that are used to represent aspects of tracked objects in video frames. Object modeling is done during the task of tracking previously unseen selected objects, and both tracking and model creation are implemented using the WiSARD weightless neural network and occur in real time, starting from no prior knowledge. The main purpose of this work is to track an object through camera images and, simultaneously, create a model that describes the presented appearances along with the transitions between each learned aspect. To achieve this goal, an object tracker based on the ClusWiSARD weightless neural network model was used to determine the states that describe the observed objects. In this way, it is possible to obtain a system that capture knowledge about the visual structures of the learned objects, creating relationships between the possible appearances, and being able to transit over the model aspects in an appropriate way. Furthermore, the created models have visual representations that can be used to show the learned aspects and validate the state transitions, in addition to being able to visualize occluded parts of objects. [ABSTRACT FROM AUTHOR]